CN114096194B — Systems and methods for cognitive training and monitoring

Assignee: Aceral Co Ltd · Inventors: Yair Gilutz (耶尔·吉卢茨), Shai Granot (沙伊·格拉诺特), Anna Izotzky (安娜·伊左特奇夫) Filed: 2020-05-04 (CN202080040282.1A; US17/615,200) · Granted: CN114096194B, 2022 · Status: Active · Anticipated expiration: 2040-05-04

Abstract

A system and method for analyzing user feedback in response to a cognitive training program. Machine learning algorithms predict training success rate using predefined datasets of user feedback. New user feedback is received, prediction determined, and the cognitive training program modified accordingly. The ML is retrained with reinforcement learning. Behavioral patterns are monitored, and alarms issued if training success rate decreases beyond a threshold.

Key technical features

• Brain recording: OPTIONAL — EEG sensor (claim 18); eye-tracking imager (claim 22); ML (LSTM/RNN, G06N3/0442) for training success prediction; 23 EEG mentions; 1 “brain wave” mention
• Species: Human — 9 “human” mentions; 212 “user”; 2 “patient”; zero animal/rodent/primate
• Classification: A61B5/16 (psychotechnics); A61B5/163 (eye movement/gaze/pupil); A61B5/4088 (cognitive diseases — Alzheimer/dementia); G16H20/70 (mental therapies); G06N3/0442 (LSTM); G06N3/088 (unsupervised learning); G06N3/09 (supervised learning)
• ML approach: Recurrent neural networks (LSTM/GRU); supervised + unsupervised + reinforcement learning
• Monitoring: Timing, training session length, success rate, attention stability, freeze period, number of breaks
• User profiling: Gender, age, education, location, language, occupation, marital status

Claims (selected)

• Claim 13: System with processor storing ML algorithm, training data, user feedback; predicts training success rate; modifies cognitive training program; retrains with reinforcement learning; monitors behavioral patterns; issues alarm on decreased success rate.
• Claim 18: EEG sensor coupled to processor; user profile determined based on measured EEG signals.
• Claim 19: ML algorithm implemented on recurrent neural network with LSTM.
• Claim 22: Imager coupled to processor; monitors eye movement of user.

Significance

A cognitive training + monitoring system from Aceral (Israeli company). The core method is behavioural (user feedback to cognitive training exercises) with optional EEG and eye-tracking as supplementary monitoring. The ML pipeline (LSTM + reinforcement learning for adaptive training) is sophisticated. Targets cognitive diseases including Alzheimer/dementia (A61B5/4088 classification). The EEG is not the primary modality — it is an optional sensor for user profiling — so this patent sits at the boundary between behavioural and brain-recording cognitive assessment.